摘要
针对目前人脸图像处理中比较热门的笑脸识别领域,本文以几何特征和纹理特征为基础,首先对人脸图像进行预处理,找到人脸部分,对人脸部分使用ASM算法提取几何特征,然后通过LBP算法提取纹理特征,使用加权量化的方法进行融合,最后将融合后的特征送入SVM中进行表情识别分类。本文在JAFFE表情数据库上进行了算法验证和比对,实验结果表明使用多特征融合比使用单一特征在人脸表情识别率上有较大的改善。
Aiming at the popular smiley face recognition field in face image processing,based on geometric features and texture features,this paper first preprocesses the face image to find the face part,and then uses ASM algorithm to extract the geometric features of the face part,and then uses LBP algorithm to extract the texture features,and uses weighted quantization method for fusion,Finally,the fused features are fed into SVM for expression recognition and classification.In this paper,the algorithm is verified and compared on Jaffe expression database.The experimental results show that the recognition rate of facial expression is greatly improved by using multi feature fusion than using single feature.
作者
王涛
彭欣荣
朱经睿
李威
Wang Tao;Peng Xinrong;Zhu Jingrui;Li Wei(Ji’an vocational and Technical College,Ji’an Jiangxi,343000)
出处
《电子测试》
2021年第23期52-54,共3页
Electronic Test
基金
吉安职业技术学院校级自然科学课题“基于LBP和几何特征融合的笑脸特征提取算法的研究”。
关键词
表情识别
几何特征
特征融合
ASM
LBP
expression recognition
Geometric features
Feature fusion
ASM
LBP
